Almost every business owner opening their Trustpilot dashboard for the first time misreads the TrustScore as a straight average of their star ratings. It is not, has not been since 2019, and the difference matters — because the four hidden weightings determine whether posting 40 new reviews next month raises your score by 0.1 or 0.5, and whether removing one fake one-star nudges the number visibly or not at all.
I am Robiul, head of research at BGR Review. Our team models TrustScore movement for clients every week and the numbers below are drawn from Trustpilot's own documentation, its 2024 methodology update and pattern data from client profiles we have monitored across 2024 to 2026.
The four weightings that actually move the TrustScore
Trustpilot does not publish the exact coefficients but has publicly documented the four inputs. Understanding what each one does is the difference between chasing the wrong lever for six months and moving the score meaningfully in one quarter.
1. Recency weighting
Recent reviews count more than old ones. A 5-star from last week has materially more impact than a 5-star from three years ago, and the decay is gradual rather than a hard cliff. This is why a profile with 2,000 lifetime reviews at an average of 4.5 can display a TrustScore of 3.8 if the last 90 days skew heavily negative — the old positives are still counted, just underweighted. The practical implication is that a consistent flow of fresh reviews matters more for the visible score than the size of the historical archive.
2. Bayesian pull toward the platform mean
For profiles under roughly 100 reviews, the TrustScore is pulled toward the platform's overall mean of approximately 3.5. This is standard Bayesian smoothing and prevents a single 5-star review from producing a 5.0 score, or two 1-stars from producing a 1.0. As review count grows past 200, the pull weakens materially; past 1,000 it is negligible. This is the single biggest reason small profiles feel harder to move — the maths is deliberately conservative until you have enough sample size to be trustworthy.
3. Anti-manipulation dampener
Unusual activity — a sudden burst of 5-stars, a coordinated 1-star cluster, IP or device anomalies — triggers a temporary dampener that reduces the weight of the suspicious reviews while the integrity team investigates. Reviews flagged as fake and subsequently removed are backed out of the historical calculation, so removals can shift the score noticeably weeks after the flag was filed. The detection stack behind this dampener is documented in our fake-review spotting guide.
4. Symmetric star weighting
One 5-star and one 1-star cancel each other almost exactly in the calculation — the extremes carry equal weight in opposite directions. A 4-star has roughly half the upward pull of a 5-star, a 2-star roughly half the downward pull of a 1-star, and 3-stars sit near neutral. This is why a profile can show a stable TrustScore of 4.3 while receiving both effusive 5-stars and angry 1-stars each week — the two poles are balancing.
The most common client question we get is 'why did the TrustScore drop when we removed a fake 1-star?' The answer is almost always that the removal shifted the recency window and revealed a slower underlying trend the fake had been masking.
TrustScore vs star rating vs Stars category
Three related numbers appear on every Trustpilot profile and they are not the same thing. Confusing them is the reason many owners chase the wrong metric.
- TrustScore: the weighted 0-5 numerical score (displayed to one decimal) — the primary trust signal, calculated from the four weightings above.
- Star rating: a 1-5 star visual derived from the TrustScore band — 4.5 or higher is 'Excellent' (5 stars), 4.0 to 4.4 is 'Great' (4 stars), 3.0 to 3.9 is 'Average' (3 stars), 2.0 to 2.9 is 'Poor' (2 stars), 0 to 1.9 is 'Bad' (1 star). Two profiles with TrustScores of 4.5 and 4.9 both display 5 stars visually.
- Individual review star: the 1-5 the reviewer gave. Feeds into the TrustScore calculation via the symmetric weighting above, but is not itself the TrustScore.
How many reviews does it take to move the TrustScore?
Model output from client-profile monitoring, treating each scenario as a starting point rather than a fixed rule. Every profile is different — the recency window, existing volume and star distribution all shift the answer.
- Under 50 total reviews: Bayesian pull is strong. Adding 20 five-star reviews typically moves a 3.8 score to roughly 4.1-4.2, not 4.7. Adding 5 one-stars typically drops it to 3.5-3.6, not 3.2. Small numbers, big pull toward the mean.
- 200-500 total reviews: Bayesian pull is moderate. Adding 40 fresh five-stars over a month typically moves a 4.1 to roughly 4.3-4.4. This is the range where consistent invite-driven capture starts to show visible weekly progress.
- 1,000+ total reviews: Bayesian pull is minimal but recency still dominates. Moving a score from 4.3 to 4.5 typically requires 150-250 fresh five-stars over 60-90 days, or aggressive removal of aged 1-stars via guideline appeals.
- Any size, after a Consumer Warning banner: assume the visible score is depressed by 0.3-0.6 regardless of underlying reviews, and cannot fully recover for 12 months while the banner is active.
The three legitimate levers that actually raise the TrustScore
Only three moves consistently raise the number without breaching guidelines. Everything marketed as a shortcut usually falls into one of these three or is illegal — see our legal breakdown of buying Trustpilot reviews for what to avoid.
- Automate a post-purchase invite inside 24 hours of the transaction. Response rates run 8-15% and halve every 48 hours of delay. This is the single highest-ROI move because the reviews it produces are Verified, recent, and additive to the recency window.
- Reply publicly to every 1-star inside 24 hours with a specific fix. Trustpilot data shows 41% of unhappy reviewers update or remove their review after a 24-hour substantive reply versus 4% past a week. Updated 1-stars re-enter the calculation with the new star, moving the score directly.
- File guideline-based removal on genuinely fraudulent negatives with a specific guideline citation (4.1 to 4.6) and evidence. Successful removals back the fake review out of the historical calculation, and the resulting shift is often larger than expected because the fake was suppressing the score via the recency weighting.
Q.How is the Trustpilot TrustScore calculated?
The TrustScore is a Bayesian-weighted average of all reviews on the profile, calculated on a 1-5 scale to one decimal place. Trustpilot applies four weightings: recency (recent reviews count more), Bayesian pull toward the platform mean of 3.5 for smaller profiles, an anti-manipulation dampener on unusual activity, and symmetric star weighting where 5-stars and 1-stars carry equal opposite impact.
Q.Why is my TrustScore different from my average star rating?
Because the TrustScore is weighted, not a simple mean. Recent reviews count more, small profiles are pulled toward 3.5, and suspicious clusters are dampened while investigated. A profile with a straight 4.5 average can display a TrustScore of 4.2 if the last 90 days skew negative, or 4.6 if the recent activity is disproportionately positive.
Q.How often does the TrustScore update?
The TrustScore updates whenever a new review is published, an existing review is edited by the reviewer, or a review is removed by Trustpilot's integrity team. There is no fixed schedule — changes typically appear within minutes of the underlying review event, though the anti-manipulation dampener can delay the impact of suspicious activity by hours or days.
Q.How many reviews do I need for a stable TrustScore?
Roughly 200 reviews is where the Bayesian pull toward the platform mean of 3.5 weakens materially. Under 100 reviews the score is noticeably pulled toward 3.5. Past 1,000 reviews the pull is negligible and the recency weighting becomes the dominant driver of movement.
Q.Do 4-star Trustpilot reviews hurt my TrustScore?
Only slightly. Under Trustpilot's symmetric star weighting, 4-stars have roughly half the upward pull of 5-stars, so they are still net positive on a profile at any score under 4.5. They become net-neutral on a profile already averaging 4.5+ and cannot pull a strong profile down materially.
Q.Can I see the exact TrustScore formula Trustpilot uses?
No. Trustpilot publishes the four inputs (recency, Bayesian pull, anti-manipulation, symmetric weighting) but does not disclose the exact coefficients or decay curves. This is a deliberate anti-manipulation choice — if the exact formula were public, farms could optimise their fake reviews against it. The four documented inputs are enough to predict directional movement reliably.
The honest bottom line
The TrustScore is a weighted trust signal, not a maths average. Owners who understand the recency window, the Bayesian pull under 200 reviews and the symmetric star weighting stop chasing the wrong lever and start compounding the right ones. Consistent invite automation plus fast substantive replies to 1-stars plus disciplined guideline appeals on genuine fakes moves the number more predictably than any campaign that tries to shortcut the system — and it is the only workflow that survives Trustpilot's retro-audits with the score intact.
